4
High
|
64 bit sync primitives
|
0
Approved
|
11
Implemented
|
Dr. David Alan Gilbert
|
Dr. David Alan Gilbert
|
Michael Hope
|
4
High
|
Conditional compares
|
0
Approved
|
11
Implemented
|
Ramana Radhakrishnan
|
Ramana Radhakrishnan
|
Michael Hope
|
4
High
|
Fix widening multiply scheduling
|
0
Approved
|
11
Implemented
|
Ramana Radhakrishnan
|
Ramana Radhakrishnan
|
Michael Hope
|
4
High
|
Forward port the patches from Linaro GCC 4.5 to 4.6
|
0
Approved
|
11
Implemented
|
|
Michael Hope
|
Michael Hope
|
4
High
|
Backport any Tuning for other ARM cores from upstream
|
0
Approved
|
11
Implemented
|
Ramana Radhakrishnan
|
|
Michael Hope
|
4
High
|
Improve constant generation in Thumb-2
|
0
Approved
|
11
Implemented
|
Andrew Stubbs
|
Andrew Stubbs
|
Michael Hope
|
4
High
|
Make -fsched-pressure a more consistent win
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Richard Sandiford
|
Michael Hope
|
4
High
|
Check common programs for areas the vectoriser could improve
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
4
High
|
Linaro GCC the product
|
0
Approved
|
11
Implemented
|
Andrew Stubbs
|
Michael Hope
|
Michael Hope
|
4
High
|
NEON performance for 11.11
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
4
High
|
Thumb-2 performance for 11.11
|
0
Approved
|
11
Implemented
|
|
Ramana Radhakrishnan
|
Michael Hope
|
4
High
|
Fix any NEON vs core regressions
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
4
High
|
Reduce over-promotion
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
4
High
|
Reduce redundant VFP/ARM moves
|
0
Approved
|
11
Implemented
|
Ramana Radhakrishnan
|
Ramana Radhakrishnan
|
Michael Hope
|
4
High
|
Improve SLP data analysis
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Richard Sandiford
|
Michael Hope
|
4
High
|
Improve SMS on code with memory dependencies
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
4
High
|
Track and investigate performance regression areas for GCC
|
0
Approved
|
12
Informational
|
|
Ramana Radhakrishnan
|
Michael Hope
|
4
High
|
Improve support for widening multiplies for ARM
|
0
Approved
|
11
Implemented
|
Andrew Stubbs
|
Andrew Stubbs
|
Michael Hope
|
4
High
|
Tidy vectoriser testsuite
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
4
High
|
Vectoriser doubling multiply support
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
4
High
|
Initial bring up so that toolchain can debug issues found via Android
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
4
High
|
Investigate libav inline assembly
|
4
Discussion
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
3
Medium
|
64 bit divide by constant
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
3
Medium
|
64 bit ones complement in NEON
|
0
Approved
|
11
Implemented
|
Andrew Stubbs
|
Michael Hope
|
Michael Hope
|
3
Medium
|
64 bit operations
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Michael Hope
|
Michael Hope
|
3
Medium
|
64 bit twos complement in NEON
|
0
Approved
|
11
Implemented
|
Andrew Stubbs
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Backport A15 branch tuning
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Backport conditional execution work
|
0
Approved
|
12
Informational
|
|
Andrew Stubbs
|
Michael Hope
|
3
Medium
|
Backport double compare
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Backport initial A7 support
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Backport 64 bit multiply, divide, and mod improvements
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Better 64 bit load immediate
|
0
Approved
|
11
Implemented
|
Andrew Stubbs
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Better 64 bit shifts in core registers
|
0
Approved
|
11
Implemented
|
Andrew Stubbs
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Better 64 bit shifts in NEON
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Better extending core to NEON transfers
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Fix EPILOGUE_USES regression in CoreMark
|
0
Approved
|
12
Informational
|
Kugan Vivekanandarajah
|
Ramana Radhakrishnan
|
Michael Hope
|
3
Medium
|
Generic tuning for all Cortex-A devices
|
0
Approved
|
12
Informational
|
|
Ramana Radhakrishnan
|
Michael Hope
|
3
Medium
|
Improve constant pool support
|
0
Approved
|
12
Informational
|
|
Ramana Radhakrishnan
|
Michael Hope
|
3
Medium
|
Improve IV opts #1
|
0
Approved
|
12
Informational
|
|
Ramana Radhakrishnan
|
Michael Hope
|
3
Medium
|
Improve peeling
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
3
Medium
|
Improve the register choice in the allocator
|
0
Approved
|
12
Informational
|
|
Ramana Radhakrishnan
|
Michael Hope
|
3
Medium
|
Native tuning
|
0
Approved
|
11
Implemented
|
|
Andrew Stubbs
|
Michael Hope
|
3
Medium
|
Improve extraction of double vectors from quad vectors
|
0
Approved
|
11
Implemented
|
Richard Sandiford
|
Richard Sandiford
|
Michael Hope
|
3
Medium
|
NEON instruction coverage
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Improve extraction of vectors from strided loads
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Richard Sandiford
|
Michael Hope
|
3
Medium
|
NEON widening left shift
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
3
Medium
|
Profiler driven feedback
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Ulrich Weigand
|
Michael Hope
|
3
Medium
|
Make SLP try different vector widths
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Richard Sandiford
|
Michael Hope
|
3
Medium
|
Implement SMS improvements for 1111
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Ramana Radhakrishnan
|
Michael Hope
|
3
Medium
|
SMS: support auto-inc
|
0
Approved
|
11
Implemented
|
Revital Eres
|
Revital Eres
|
Michael Hope
|
3
Medium
|
Support VCVT for fixed point/float
|
0
Approved
|
11
Implemented
|
Ramana Radhakrishnan
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Linaro toolchain in Android
|
0
Approved
|
11
Implemented
|
Ken Werner
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Unaligned struct access
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Use vld1 for quad loads
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Widen default vector size
|
0
Approved
|
11
Implemented
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
3
Medium
|
Write up building at O3
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Investigate IV opts
|
4
Discussion
|
12
Informational
|
|
Richard Sandiford
|
Michael Hope
|
3
Medium
|
Investigate NE10
|
4
Discussion
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Allow more operations to be SLPed
|
4
Discussion
|
11
Implemented
|
|
Richard Sandiford
|
Michael Hope
|
3
Medium
|
Make SMS schedule the register moves that it creates
|
5
New
|
11
Implemented
|
Richard Sandiford
|
Richard Sandiford
|
|
3
Medium
|
Better 64 bit multiply and divide
|
6
Superseded
|
5
Started
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
3
Medium
|
Choose vector width automatically in SLP
|
6
Superseded
|
5
Started
|
Ira Rosen
|
Ira Rosen
|
Michael Hope
|
3
Medium
|
Improve sched-pressure for ARM
|
6
Superseded
|
11
Implemented
|
Ulrich Weigand
|
Michael Hope
|
Michael Hope
|
2
Low
|
Linaro GCC 4.6 in Distributions
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
2
Low
|
Add multiply pipeline bypass
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
2
Low
|
<arm_neon.h>/intrinsics improvements
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
2
Low
|
Improve vectoriser narrowing operations
|
0
Approved
|
12
Informational
|
|
Ira Rosen
|
Michael Hope
|
2
Low
|
Deeper toolchain validation
|
0
Approved
|
11
Implemented
|
|
Michael Hope
|
Michael Hope
|
2
Low
|
Reduce use of VMOV
|
0
Approved
|
11
Implemented
|
Ulrich Weigand
|
Richard Sandiford
|
Michael Hope
|
2
Low
|
Add backend support for NEON's VABD
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
2
Low
|
Linaro toolchain in Ubuntu
|
0
Approved
|
11
Implemented
|
|
Michael Hope
|
Michael Hope
|
2
Low
|
Unaligned memcpy()
|
0
Approved
|
11
Implemented
|
Michael Hope
|
Michael Hope
|
Michael Hope
|
2
Low
|
Better multilib defaults
|
4
Discussion
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
0
Not
|
ARMv5 saturating add/subract support
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
|
0
Not
|
Add ARMv6 SIMD support
|
0
Approved
|
12
Informational
|
|
Andrew Stubbs
|
Loïc Minier
|
0
Not
|
Improve block memory operations by GCC
|
0
Approved
|
12
Informational
|
|
Ramana Radhakrishnan
|
Michael Hope
|
0
Not
|
Transform statics to locals
|
0
Approved
|
12
Informational
|
|
Michael Hope
|
Michael Hope
|
0
Not
|
Investigate whether peeling is effective for neon
|
6
Superseded
|
1
Not started
|
|
|
Michael Hope
|